Skip Navigation


20 seconds spare? Run a speed test
Thursday 19 May 2016 10:23:34 by Andrew Ferguson

Our speed tests have a tendency to get more and more complex as everyone asks for various features to be added, but we thought lets do a stripped back version for once.

Our regular visitors will be well aware of our monthly speed test summary which provides regular and timely updates on what people are actually experiencing rather than what the PR machines of the providers want us to believe. All the speed tests with a postcode feed into our coverage site where the postcode level analysis updates monthly and also features a more complex speed tester that lets you browse other peoples tests once completed.

Comments

Posted by Blackmamba 7 months ago
Hi Andrew Staff
A much better speed test just run three on Gu266qd and it is showing on my upgrade from 40/10 to 50/10 that the average is increasing could this be due to the DLM kicking in time will tell.
Posted by Llety 7 months ago
2 out of 3 times our ADSL line times out on a speed test (any speed test).

Does this get recorded as 0 and added into the mix?
Posted by andrew (Favicon staff member) 7 months ago
If the test makes it to the results page it has been recorded by us as a valid test. We do record failures but impossible to know reasons for failure so are not added to the overall statistics.

http://tbb.st/1463659182261555655 a recent slow test, and people many slower and have seem some that say 0 down and up to 1 decimal place, i.e. are below 0.05 Mbps dial-up speeds
Posted by tombartlett 7 months ago
Are there any plans to stop using port 80 so my web filter and antivirus can't interfere with the results?
Posted by andrew (Favicon staff member) 7 months ago
https://labs.thinkbroadband.com/speedtest Page and testing is over SSL so port 443

Posted by Llety 7 months ago
I agree you can't know the reason for failure, but if a download starts and does not finish, the most likely explanation is network failure and so should be considered when counting the average for the post code or at least sample size and number of failures reported.

Of course the submitter could have got bored, laptop run out of power, etc, etc.

Not factoring in failed speed tests as 0 skews the numbers away from the real experience.
Posted by francisuk20 7 months ago
Here is a speedtest on Three 4G on 1 bar in the HA8 area (north West London)
http://www.thinkbroadband.com/speedtest/results.html?id=1463663146582947255
Posted by andrew (Favicon staff member) 7 months ago
The problem is we do not know if the user would have recorded 0.00001 Mbps, 0.1 Mbps or 10 Mbps or 100 Mbps so what do you record?

Posted by Apilar 7 months ago
Are you going to ditch the flash speed test at some stage?
Posted by Llety 7 months ago
Record that attempt as zero. If you have 2 samples, 1 worked at 10mb, the other failed after the session started, then the mean is 5mb.

I believe it is the right thing to do from a statistical perspective, unless you also show the
number of success and failures.

Posted by tommy45 7 months ago
@ Apilar: I hope they don't Because for me the flash tester gives more reliable results, the HTML5 tester seems to be too fussy about which browser is being used and AV security suite in use, On my Win 7 x64 box there's a bug on both speed testers where the single thread test will max at around 28-30mbps opera and HTML5 is quite persistent but randomly does this with the flash based tester, Firefox displays issues too
Posted by ian72 7 months ago
@Llety they shouldn't be recorded at all as a failed test has no bearing on the line speed as it could have been the browser crashing or similar. Someone looking to understand line speeds does not want duff tests that have possibly got nothing to do with the connection itself.
Posted by ian72 7 months ago
Sorry, when I say not recorded I mean not included in the mean. Recording them is fine as long as it doesn't skew the stats for unknown reasons.
Posted by WWWombat 7 months ago
@llety
It might be valid to hold 5Mb as *your* average. It gives an accurate depiction of your experience.

However, reporting that as an accurate indication of what the local area can expect is simply bogus.

Why on earth would I want to see a figure that incorporates results of one person's incompetence at maintaining their PC/network/wiring/device?
Posted by WWWombat 7 months ago
Hmmm. Recently, the standard "flash" test has taken to going askew, especially upstream - which it measures as 10Mbps. Downstream isn't immune either.
http://www.thinkbroadband.com/speedtest/results.html?id=146367347525378714428

This "quick" test seems more accurate, with an 18Mbps upstream...
http://www.thinkbroadband.com/speedtest/results.html?id=1463669875288937455
Posted by WWWombat 7 months ago
@llety followup
Thinking about it further...

If you really wanted to keep some idea of tests that failed in an area, it would be far better to track the rate of success/failure separately from the actual speed average.

eg 25Mbps average, 99.3% tests completed.
Posted by tommy45 7 months ago
This 20second test has issues with Opera running on WIN 7x64 it gives a ridiculously low upload result for me less than 2mbps ( average: 0.93 and 1.05 burst) yet firefox on the same box gives a more realistic result on the upstream of 18.2mbps and 19.2 burst,The downstream results with firefox where more accurate, as was the latency of 26ms, as opposed to the opera result of 170ms
Posted by andrew (Favicon staff member) 7 months ago
@tommy45 So this is different behaviour to http://labs.thinkbroadband.com/speedtest and https://labs.thinkbroadband.com/speedtest is that what you are saying?
Posted by tommy45 7 months ago
From the testing i have done so far it's mainly the upstream being lower with this 20sec test, although the last i did with firefox on the xp box showed an improvement though still a bit low
Posted by tommy45 7 months ago
http://www.thinkbroadband.com/speedtest/button/1463695552725357855.png
above link is the result from win 7 box running opera
http://www.thinkbroadband.com/speedtest/button/1463695678278137855.png is same pc but with firefox


Posted by tommy45 7 months ago
I just fan the main HTML5 test again same box os and using opera http://www.thinkbroadband.com/speedtest/button/1463696254452459355.png

same/similar low upstream results as the short test
Posted by andrew (Favicon staff member) 7 months ago
So not new test version breaking it, but something Opera is up to that needs to have a work around found.

Opera does seem to have a history of breaking uploads of material from browsers though
Posted by WWWombat 7 months ago
My upstream issue was with a chrome browser on Win7.
Posted by danielmec 7 months ago
I note the test says "Please note that this test will not link to your profile.", which is the same as running from the Big Orange Button link (actually why I visited today). Are there plans to fix this? I've found that having a historical graph of results to be really useful, especially to confirm a couple of line-faults I've had.
Posted by drwright 7 months ago
Why does the Share button take one to Faeces Book
Posted by andrew (Favicon staff member) 7 months ago
The f share button has the usual facebook share icon on it, there is also a twitter and a standard URL button.
Posted by andrew (Favicon staff member) 7 months ago
@danielmec The tester is all ready to link to the profiles it just needs some stuff done by others as part of a wider set of changes to happen.
Posted by danielmec 7 months ago
@andrew Thanks! As always its great to see tbb making usability improvements.
Posted by richi 7 months ago
Hangs at 100% download. Chrome 50.0.2661.102 m on Win10 64bit 1511
Posted by richi 7 months ago
Sorry, ignore that. PEBKAC.
Posted by andrew (Favicon staff member) 7 months ago
Okay the Opera issue is down to its turbo mode, turn it off and all is fine.
Posted by rtho782 7 months ago
I notice that your speedtests are multithreaded, this means that by load balancing I use both lines and achieve about 85mbit.

I do wonder if some poor soul coming to the area later will think that they can get that, only to discover a single FTTC line is about 45mbit...
Posted by andrew (Favicon staff member) 7 months ago
Opera users now have a warning message - have tried to detect turbo mode but it seems Opera changes its mind sometimes, so you can be part way through a test and flips, so not reliable enough as a warning before hand.
Posted by andrew (Favicon staff member) 7 months ago
On load balancers - rare enough to not be a major issue
Posted by danieltemple 7 months ago
This speediest is nonsense. I average 71Mb/s on various speed test services. This one shows 14Mb/s.
Posted by andrew (Favicon staff member) 7 months ago
@danieltemple Sorry to hear that, can you visit http://labs.thinkbroadband.com/speedtest and run that test and post a copy of the share results link at the end so we can look at why this may be.

Posted by chrysalis 7 months ago
why is this threaded only? lets poor performing providers slip through.
Posted by rndmartin 7 months ago
Is this one still based on flash?
Posted by andrew (Favicon staff member) 7 months ago
No - no flash in this or the version at http://labs.thinkbroadband.com/speedtest or on https://labs.thinkbroadband.com/local (in the search tab)
Posted by hairybarsteward 7 months ago
How accurate is this version of the test? I only ask as I log into my account and get 73Mbps Down and approx 11 Mbps up on the speed test. My wife can switch profiles, go to your site, runs a test without logging into an account and gets 75Mbps Down and 18.5Mbps up on the speed test. Same machine only 30 seconds difference between tests. Any ideas?
Posted by andrew (Favicon staff member) 7 months ago
@hairybarsteward Believe overall the new test (that has been running for 3 years, this is just a different skin) is the best and this gave you http://tbb.st/1463749053782821855 72Mbps down 18Mbps up

The older flash based test gave a slower upload http://tbb.st/146288941967770526637 and this is why its slowly being phased out.
You must be logged in to post comments. Click here to login.